Aiming at the characteristic of salient region detection,pixel genetic algorithm is proposed.Firstly,the genetic units are mapped to image region in the genes model.Then the genes is classified by using gene expression,expectation value is counted for the uncertain information entropy,and the uncertainty ability that characteristic values reduce system entropy is calculated.Finally,salient regions and non salient regions are divided by using the adjacency matrix statistic,which the gene distribution is symmetrical.Simulation results show that the algorithm can sufficiently extract image features,and has a higher retrieval rate and accuracy.